Nathan Lambert from AI2 and Kevin Xu lay out why clamping down on open-source AI models could clash with longstanding American habits around sharing knowledge, letting newcomers tinker, and keeping any single company from charging whatever it wants.
Banning open-source AI in any form would be a mistake. A general audience PSA with @kevinsxu on why open source upholds American values.
Managing frontier risks is hard, but reducing transparency, innovation, and education from kneecapping the open frontier would be worse.
http://x.com/i/article/2067652729913114624
The piece points out that open weights already underpin a lot of current AI work and historically delivered trillions in value by lowering the bar for experiments and new startups.
Lambert and Xu flag recent policy signals as possible entry points for wider rules, yet they leave open how any actual restrictions would be scoped or enforced.
Positive users defend open-source AI against bans and export controls to uphold liberty and block gatekeepers, while negative users dismiss it as un-American and greedy.
No Digg Deeper questions have been answered for this story yet.
@kevinsxu Also on @interconnectsai and interconnected (yes, overlap)
Banning Open Source AI Would Be A Mistake A general audience PSA with Kevin Xu of Interconnected https://www.interconnects.ai/p/banning-open-source-ai-would-be-a
> Without the equalizing and democratizing nature of open source, we would all be living with the rent-seeking consequences of more monopolies and less free market competition.
🎯
Banning open-source AI in any form would be a mistake. A general audience PSA with @kevinsxu on why open source upholds American values.
Managing frontier risks is hard, but reducing transparency, innovation, and education from kneecapping the open frontier would be worse.
Banning open-source AI in any form would be a mistake. A general audience PSA with @kevinsxu on why open source upholds American values.
Managing frontier risks is hard, but reducing transparency, innovation, and education from kneecapping the open frontier would be worse.
@natolambert @kevinsxu the more time goes on the more the depth of this sinks in. they are setting up testing and criteria all models will have to go through, we are on hold until that happens. think we might have truly crossed the line: real regulation, possibly heavy, and things are different now.
@xeophon hey i think this might be a lot more nasty than it appears. we may really have crossed over into new territory, every meaningful release will be fable+ from here on out and the [small] models will start hitting the ceiling soon.
it's the testing, can't see a clean way that works
@natolambert @kevinsxu Shut up! In China, everything is open source, from mommy's secret sauces, to hot pop songs. Everybody just takes and uses, never wants to pay. How dare are you saying Open Source upholds American values? Objection!
@natolambert @kevinsxu Open source can't uphold any American value (e.g. psychopathy, endless greed)
@natolambert @kevinsxu In China, everything is open source, from mommy's secret sauces, to hot pop songs. Everybody just takes and uses, never wants to pay.
How dare are you saying Open Source upholds American values?
Objection!
@natolambert @kevinsxu Liberty must prevail.
@natolambert @kevinsxu Useful if it changes what builders or buyers do next.
@natolambert @kevinsxu But almost none of the open-weight models on the market are open-source. The data used to train them is not shared (and is often legally suspect) so you can't replicate them on your own; you can only take what is given to you by some big company. What was that about transparency?
@natolambert @kevinsxu Every time "frontier risks" comes up it's the same crowd that benefits from keeping everything locked down. @SentientAGI are proving you can build transparent infrastructure without the gatekeeping nonsense
@AndrewCurran_ Very much a genie out of the bottle situation anyway. If open weight modes are banned, only criminals will have powerful AI.